Short-term Load Forecasting Based On Wavelet

As an integral part of power system, short-term load forecasting has an important impact on the power system security and economic development. Short-term load forecasting research has a long history, the researchers continues to put forward to methods that may improve the accuracy of load forecasting. In recent years, the data mining, machine learning, artificial intelligence technology and some other technologies are more and more widely used in electric power system load forecasting, the role of the theory of wavelet analysis in power system load forecasting becomes more and more significant. Gradient boosting regression tree algorithm has a broad application in the search rankings, machine learning, and biological research and other fields. It is the extension of Boosting algorithm.This paper introduces the concept and significance of load forecasting, research status of short-term load forecasting, the basic principle of wavelet transform and the gradient boosting regression. Combined with the feature of meteorological data and the trend of load data, an approach to short-term load forecasting is proposed, that is the gradient boosting regression tree based on wavelet.Power load has many characteristics, such as instability, nonlinear and cyclical. Wavelet transform is a kind of time domain and frequency domain analysis method. Through the wavelet transform the sequence is decomposed into periodic subsequence, this sequence more cyclical than the original sequence, So this article will first decompose and reconstruct load sequence and meteorological sequence respectively, then build a model on refactoring load subsequence and meteorological subsequence with gradient boosting regression method. The calculation results of a practical example show that the proposed method possesses high forecasting accuracy compared with other methods such as the single gradient Boosting regression, the decision tree and wavelet decision tree.